If you take today’s tradable coin list and backtest it several years into the past, the dataset can look deceptively clean. Coins that existed then but were later delisted or lost liquidity may be missing.

Today's list may contain only the survivors

Assets that are still trading today have already survived a filter. If coins that later disappeared would have hurt the strategy, a backtest using only the current list will never see those failures.

SURVIVORSHIP BIAS
Past universe→Winners + failures→Today→Survivors only

The asset universe must also be point-in-time data

Use the assets that were actually tradable on each historical date, together with listing and delisting timestamps, to reconstruct the opportunity set that existed at the time.

Liquidity filters must not become hindsight filters.

Applying today’s high-volume filter to the past creates another form of hindsight selection. Volume and liquidity filters must be computed from information available at the historical decision time.